I have a job that forces me to learn constantly. As an IT infrastructure and operations leader, there is a new tool, a new vendor, a new security model, or a new certification on my plate basically every week. For about a year now, the tool I reach for to get up to speed fast is Google’s NotebookLM. It has quietly become one of the most useful things in my stack — not because it is flashy, but because it changed how I learn.
This is not a feature tour you could get from Google’s marketing page. This is exactly how I use it, what it has actually done for my career and my life, and where it still falls short.
What NotebookLM actually is
NotebookLM is a source-grounded research and learning tool. You feed it your own sources — PDFs, web pages, Google Docs, YouTube videos — and it works from those sources instead of guessing from the open internet. Then a "Studio" panel turns that material into different learning formats: Audio Overviews (podcasts), Video Overviews, Mind Maps, Data Tables, Quizzes, Flashcards, and Reports. (DigitalOcean overview, Google on flashcards/quizzes)
On price: the free tier is genuinely generous — 100 notebooks, 50 sources per notebook, and 3 audio overviews a day. I used the free version hard for a long time and never hit a wall. I am now on the Pro tier (Google AI Pro, $19.99/month), but honestly I upgraded for the broader Google AI bundle, not because free ran out on me. One quirk worth knowing: you cannot buy NotebookLM by itself anymore — it ships as a benefit of a Google AI plan. (NotebookLM pricing breakdown)
My workflow, step by step
This is the part that matters. The tool is only as good as how you use it, and I have a system.
1. Gather as many sources as possible
Learning quality tracks directly with source quality. I pull from everywhere: YouTube, web research, and — my favorite trick — I use Perplexity to generate a clean, copy-paste-ready list of sources on a topic, then drop those straight into NotebookLM. For YouTube, I use a free Chrome extension called YouTube to NotebookLM that drops a one-click “NotebookLM” button right into the YouTube player, so I can push an entire playlist into a notebook at once, or grab videos one by one.
Perplexity finds the sources; NotebookLM learns from them. They are complementary, not competitors.
2. Audio Overviews — my real strategy
This is my most-used feature, and I do not just hit the button. I run a Deep Dive, long format, and I give it a goal prompt every time. Roughly: "I want to learn this for work. Use real-world examples and analogies that make it easy to understand, and build me up from beginner to expert through the episode." That instruction makes it carry me from the very start of a topic all the way into the corporate-level nitty-gritty.
The second move is just as important: I split big topics into smaller ones and go deep on each. When I was learning Azure AI, I did one dive purely on voice AI — the different models, how it charges, how you administer it — then moved to the next component in its own dive. Going narrow and deep is what makes it stick.
I treat the audio like a podcast. Dishes, taking out the trash, cleaning the car, the commute, driving my wife around — that’s all learning time now.
That is the real unlock. The audio fits into the dead time of my day, so I am learning while I would otherwise be doing nothing.
3. The Data Table — my sleeper MVP
Everyone talks about the podcasts. For me, the Data Table is the underrated hero. I have it build one big table of every question and answer so I can read the Q and the A down a single line for a fast refresh. Then I build a second table of the key definitions to cement and pin those in my brain. It is the fastest review tool in the whole app.
4. Quizzes — three difficulty passes
I deliberately generate quizzes in three levels. First an easy one — questions I could almost answer with my eyes closed. Then a medium one, where I have to actually know the material to get it right. Then a hard one: "give me the questions people most often miss." Climbing those three rungs tells me exactly where my gaps are.
5. Flashcards — useful, but be honest
Flashcards are fine. The flip-to-reveal motion is good for learning through the effort, but it is slow and a little tedious. For a quick read, the Data Table beats them every time. I use both, but I lean on the table.
Does it actually work?
My test for any learning tool is simple: can I speak to the topic, or does it change my life? NotebookLM passes on both.
At work, I have become the go-to person on tools like AI Foundry and Copilot Studio, because I am constantly learning their ins and outs — and I bring real-world examples I picked up from the podcasts straight into meetings. When my company was evaluating ServiceNow’s AI Control Tower, I gathered every source I could, built a podcast explaining how it works and the security concerns it addresses, and made a Data Table breaking down its AI security features, enterprise benefits, integrations, how you control it, and whether it is actually implementable. That table did real work for a real decision. I have done the same to compare SIEM/XDR ecosystems — Palo Alto, Trellix, and Microsoft, side by side, strengths and weaknesses.
I get incredibly targeted information. That’s the difference between this and a generic chatbot.
And it is not just work. I have used it to study for certs, build executive presence for meetings, get smarter on 1031 exchanges and taxes for my real estate business, plan camping trips (a Nevada YouTuber’s whole channel became one notebook), research a possible trip to Norway, and even learn how to blog — which helped build this very site.
The honest downsides
No tool is perfect, and you come here for the warts.
My biggest gripe: there is no easy way to pipe NotebookLM’s output into other AI tools. Right now, if I want another tool to use what NotebookLM produced, I have to prompt it for the artifact, save it out, and hand it over manually. There is no clean connection. My understanding is Google is building cross-tool integration inside its own ecosystem, but I do not know if they will ever open it up beyond Google.
Beyond that: flashcards are slower than they should be, and the free-tier limits are honestly a mystery to me because I never bumped into them.
The privacy reality (my IT hat is on)
Here is the part most reviews skip. My company has not sanctioned NotebookLM, so I keep my use strictly generic. I use it to learn about technologies — "teach me about X" — and I never feed it company data, internal documents, or anything proprietary. The only thing I give it is the prompt and the way I want to be taught. It is purely for sharpening my own professional skills until it is approved.
If you work somewhere with real data governance, do the same. Learn public topics with it; do not pour your employer’s information into an unsanctioned tool.
Worth noting: Copilot has its own notebook feature, but it is a different animal — it is anchored in your Microsoft 365 Graph data, so it cannot really chew on YouTube videos or the broad general sources NotebookLM thrives on. I will do a full NotebookLM vs Copilot Notebook head-to-head soon.
Worth it if. Skip it if.
Worth it if you have to learn new things regularly — especially if, like me, that is a couple of new topics every week. It is so fast and so easy that it genuinely reorganizes how you approach learning. Start on the free tier; it is plenty.
Skip it (for now) if you need it to act on your private or work data in real time, or to plug directly into your other tools and workflows — it is not there yet. And know that you get out what you put in: the source-gathering effort is the price of admission.
The verdict
A year in, NotebookLM is the tool that turned my dead time into learning time and made me the person in the room who actually understands the new thing. It is not a magic brain — you still have to feed it good sources and ask good questions. But if you do, it is the closest thing I have found to a personal tutor that builds exactly the lesson you need, in the format you learn best. For anyone who has to keep getting smarter for a living, it is an easy recommendation.
Reviewed by Brad Rowland — IT Infrastructure and Operations leader, automation builder, and AI implementer. Based on roughly a year of near-daily use. Sources for tool facts and pricing are linked inline.
